This package provides a ROS2 node for processing images using the Ultralytics YOLO detection model. Published an output image with colored in bounding boxes. Our usage is as a pre-processing step for a Reinforcement learning model.
To run the image_node, ensure that you have ROS2 installed and properly set up. You can then build the package and run the node using the appropriate ROS2 commands.
ros2 run ultralytics_cones image_nodeOutput topics are only published if subscribed to.
- /image_node/input: Input image topic (e.g., from a camera or video source).
- /image_node/output: Output image after processing.
- /image_node/output/grey: Output image as grayscale.
- /image_node/debug: Input image showing label and bounding box annotations.
Only available if output_res parameter set
- /image_node/output/resized: Output image resized to the specified dimensions.
- /image_node/output/resized/gray: Output image resized and converted to grayscale.
- model_path: Path to model file.
- Local file, e.g. "models/cones_yolo11n_epochs10.pt"
- If local file is not found will attempt to download the model from the Ultralytics repository.
- Default: yolov8n.pt
- output_res: Output image resolution.
- Expected format is "WIDTHxHEIGHT".
- Default: "" (no resizing)
TODO - parameter for setting colors of bounding boxes.
Demos depends on the mp42rosimg package to publish mp4 video to ROS2 image topic.
https://github.com/PINTO0309/mp42rosimg
ros2 launch ultralytics_cones demo.pyros2 launch ultralytics_cones demo_resize.py
/image_node/output/debug